Update app.py

app.py

@@ -10,7 +10,11 @@ from langchain_community.utilities import WikipediaAPIWrapper
 from langchain_community.tools import WikipediaQueryRun
 from langchain_experimental.utilities import PythonREPL
 
-llm = pipeline("text-generation", model="distilgpt2")
+llm = AutoModelForCausalLM.from_pretrained("zephyr-7b-beta.Q4_K_S.gguf",
+model_type='mistral',
+max_new_tokens = 1096,
+threads = 3,
+)
 
 class SmallerQuestions(BaseModel):
     questions = list[str]